The Site Effect in Ground Motion Variability: An Application of the Variance-Components Technique

In the last decade strong-motion data in Taiwan have been increased tremendously. The recordings from the main shock and aftershocks of the 1999 Chi-Chi earthquake enrich the database that they offer novel opportunities to explore the variability of ground motions in Taiwan. More recently shear velocities for a portion of the strong-motion accelerographs deployed island-wide are becoming available from the National Center for Research on Earthquake Engineering (NCREE) in Taiwan. A data set, comprised of 3274 records from 164 earthquakes and 170 stations where shear velocities averaged over upper 30 m (Vs30) are available, is compiled. Applying the variance-components technique (Chen and Tsai, 2002) to this data set, we perform regression analyses and decompose the prediction errors into the earthquake-to-earthquake component, the site-to-site component, and the residual. The response variables explored are peak ground acceleration (PGA) and spectral accelerations (Sa) with oscillation periods ranging from 0.075 to 6 sec. For each regression relationship, the station correction factors (bs) are compared with the corresponding Vs30 at all sites. The correlation coefficients of bs and Vs30 show that they are negatively correlated. The most apparent correlation reveals a value close to 0.700 (the absolute value) falling into spectral accelerations at periods 1.5 to 3 sec. For ground motion parameter of PGA the correlation coefficient is as low as 0.450, implying that the station correction factors obtained from regression analysis for PGA are not strongly correlated with the shear velocity values averaged over the upper 30 meters at the sites.